基于GC-MS结合偏最小二乘判别分析的薰衣草精油指纹图谱研究

摘要

建立气相色谱-质谱联用(GC-MS)技术测定薰衣草精油的方法,分析了新疆法国蓝和C-197(2)2个不同品种的21批薰衣草精油样品,并基于该方法建立了薰衣草精油的指纹图谱.以GC-MS结合保留指数对复杂未知物进行定性分析,共确定了26个共有峰.采用偏最小二乘判别分析(PLS-DA)对训练集样本进行模式识别,根据模型的变量权重系数(VIP)筛选出了对品种的分类具有较大贡献的潜在标志物.并通过PLS-DA模型对7个未知样品进行预测,2个品种的薰衣草精油的均方根预测偏差(RMSPE)均为0.1323,模型对验证集中薰衣草样本的判别准确率为100%.结果表明:该方法精密度好,简单快速,为新疆薰衣草精油的品种鉴定与质量评估提供了可靠的依据.%Aim to establish a fingerprint analysis method of essential oil in lavender for quality control by gas chromatography-mass spectrometry (GC-MS).Twenty-one samples collected in Xinjiang which belong to two different lavender varieties of French blue and C-197(2) are analyzed by GC-MS.Identification of compounds is based on the retention indices and mass library.Twenty-six peaks are selected as the common peaks in fingerprint spectrum.Samples for training set are for pattern recognition by partial least squares discriminant analysis (PLS-DA).Potential biomarkers screening is performed according to VIP value.Through the forecast of 7 unknown samples in the PLS-DA model, the root mean square prediction error of two kinds of lavender essential oil samples is 0.1323, the discrimination accuracy for the lavender varieties is 100% by PLS-DA model based on the validation set of samples.The results show that the established method could be rapid and accurate to evaluate the quality of lavender essential oil.